|
|
|
--- |
|
language: |
|
- eng |
|
license: wtfpl |
|
tags: |
|
- multilabel-image-classification |
|
- multilabel |
|
- generated_from_trainer |
|
base_model: facebook/dinov2-giant |
|
model-index: |
|
- name: DinoVdeau-giant-2024_08_28-batch-size32_epochs150_freeze |
|
results: [] |
|
--- |
|
|
|
DinoVd'eau is a fine-tuned version of [facebook/dinov2-giant](https://huggingface.co/facebook/dinov2-giant). It achieves the following results on the test set: |
|
|
|
- Loss: 0.1208 |
|
- F1 Micro: 0.8209 |
|
- F1 Macro: 0.7101 |
|
- Roc Auc: 0.8812 |
|
- Accuracy: 0.3080 |
|
|
|
--- |
|
|
|
# Model description |
|
DinoVd'eau is a model built on top of dinov2 model for underwater multilabel image classification.The classification head is a combination of linear, ReLU, batch normalization, and dropout layers. |
|
|
|
The source code for training the model can be found in this [Git repository](https://github.com/SeatizenDOI/DinoVdeau). |
|
|
|
- **Developed by:** [lombardata](https://huggingface.co/lombardata), credits to [César Leblanc](https://huggingface.co/CesarLeblanc) and [Victor Illien](https://huggingface.co/groderg) |
|
|
|
--- |
|
|
|
# Intended uses & limitations |
|
You can use the raw model for classify diverse marine species, encompassing coral morphotypes classes taken from the Global Coral Reef Monitoring Network (GCRMN), habitats classes and seagrass species. |
|
|
|
--- |
|
|
|
# Training and evaluation data |
|
Details on the number of images for each class are given in the following table: |
|
| Class | train | val | test | Total | |
|
|:-------------------------|--------:|------:|-------:|--------:| |
|
| Acropore_branched | 1469 | 464 | 475 | 2408 | |
|
| Acropore_digitised | 568 | 160 | 160 | 888 | |
|
| Acropore_sub_massive | 150 | 50 | 43 | 243 | |
|
| Acropore_tabular | 999 | 297 | 293 | 1589 | |
|
| Algae_assembly | 2546 | 847 | 845 | 4238 | |
|
| Algae_drawn_up | 367 | 126 | 127 | 620 | |
|
| Algae_limestone | 1652 | 557 | 563 | 2772 | |
|
| Algae_sodding | 3148 | 984 | 985 | 5117 | |
|
| Atra/Leucospilota | 1084 | 348 | 360 | 1792 | |
|
| Bleached_coral | 219 | 71 | 70 | 360 | |
|
| Blurred | 191 | 67 | 62 | 320 | |
|
| Dead_coral | 1979 | 642 | 643 | 3264 | |
|
| Fish | 2018 | 656 | 647 | 3321 | |
|
| Homo_sapiens | 161 | 62 | 59 | 282 | |
|
| Human_object | 157 | 58 | 55 | 270 | |
|
| Living_coral | 406 | 154 | 141 | 701 | |
|
| Millepore | 385 | 127 | 125 | 637 | |
|
| No_acropore_encrusting | 441 | 130 | 154 | 725 | |
|
| No_acropore_foliaceous | 204 | 36 | 46 | 286 | |
|
| No_acropore_massive | 1031 | 336 | 338 | 1705 | |
|
| No_acropore_solitary | 202 | 53 | 48 | 303 | |
|
| No_acropore_sub_massive | 1401 | 433 | 422 | 2256 | |
|
| Rock | 4489 | 1495 | 1473 | 7457 | |
|
| Rubble | 3092 | 1030 | 1001 | 5123 | |
|
| Sand | 5842 | 1939 | 1938 | 9719 | |
|
| Sea_cucumber | 1408 | 439 | 447 | 2294 | |
|
| Sea_urchins | 327 | 107 | 111 | 545 | |
|
| Sponge | 269 | 96 | 105 | 470 | |
|
| Syringodium_isoetifolium | 1212 | 392 | 391 | 1995 | |
|
| Thalassodendron_ciliatum | 782 | 261 | 260 | 1303 | |
|
| Useless | 579 | 193 | 193 | 965 | |
|
|
|
--- |
|
|
|
# Training procedure |
|
|
|
## Training hyperparameters |
|
|
|
The following hyperparameters were used during training: |
|
|
|
- **Number of Epochs**: 150 |
|
- **Learning Rate**: 0.001 |
|
- **Train Batch Size**: 32 |
|
- **Eval Batch Size**: 32 |
|
- **Optimizer**: Adam |
|
- **LR Scheduler Type**: ReduceLROnPlateau with a patience of 5 epochs and a factor of 0.1 |
|
- **Freeze Encoder**: Yes |
|
- **Data Augmentation**: Yes |
|
|
|
|
|
## Data Augmentation |
|
Data were augmented using the following transformations : |
|
|
|
Train Transforms |
|
- **PreProcess**: No additional parameters |
|
- **Resize**: probability=1.00 |
|
- **RandomHorizontalFlip**: probability=0.25 |
|
- **RandomVerticalFlip**: probability=0.25 |
|
- **ColorJiggle**: probability=0.25 |
|
- **RandomPerspective**: probability=0.25 |
|
- **Normalize**: probability=1.00 |
|
|
|
Val Transforms |
|
- **PreProcess**: No additional parameters |
|
- **Resize**: probability=1.00 |
|
- **Normalize**: probability=1.00 |
|
|
|
|
|
|
|
## Training results |
|
Epoch | Validation Loss | Accuracy | F1 Macro | F1 Micro | Learning Rate |
|
--- | --- | --- | --- | --- | --- |
|
1 | 0.17437300086021423 | 0.21205821205821207 | 0.7424333879451582 | 0.5175126673232894 | 0.001 |
|
2 | 0.1514047533273697 | 0.24774774774774774 | 0.7776526996039191 | 0.5912510936495889 | 0.001 |
|
3 | 0.1557399332523346 | 0.23873873873873874 | 0.7752795082305376 | 0.6203462640123141 | 0.001 |
|
4 | 0.1499096304178238 | 0.2494802494802495 | 0.7691087713115115 | 0.6112936548561337 | 0.001 |
|
5 | 0.15773828327655792 | 0.24497574497574498 | 0.7744962975718961 | 0.6316545255681125 | 0.001 |
|
6 | 0.1529887616634369 | 0.25744975744975745 | 0.7803354441211706 | 0.6220908262048482 | 0.001 |
|
7 | 0.14232446253299713 | 0.2616077616077616 | 0.7837652308220353 | 0.6318272608971183 | 0.001 |
|
8 | 0.14342056214809418 | 0.2591822591822592 | 0.7824785045129828 | 0.6268140575796306 | 0.001 |
|
9 | 0.14322087168693542 | 0.25848925848925847 | 0.7840562521179261 | 0.6406683603322132 | 0.001 |
|
10 | 0.15065954625606537 | 0.2591822591822592 | 0.7779440239394473 | 0.6350156993693012 | 0.001 |
|
11 | 0.14012028276920319 | 0.26853776853776856 | 0.7905542412977358 | 0.6442254017268965 | 0.001 |
|
12 | 0.14037516713142395 | 0.26056826056826055 | 0.7896027049873203 | 0.6412994039301575 | 0.001 |
|
13 | 0.1420680731534958 | 0.2695772695772696 | 0.7822141560798549 | 0.6359393136512833 | 0.001 |
|
14 | 0.13944004476070404 | 0.2636867636867637 | 0.7887275978034142 | 0.6459907944955716 | 0.001 |
|
15 | 0.13796783983707428 | 0.2553707553707554 | 0.7915315007683115 | 0.6554204386045119 | 0.001 |
|
16 | 0.1441228836774826 | 0.255024255024255 | 0.7857792404624779 | 0.6452554527968026 | 0.001 |
|
17 | 0.14113685488700867 | 0.26784476784476785 | 0.7904489177124567 | 0.6485416937632181 | 0.001 |
|
18 | 0.1381485015153885 | 0.26056826056826055 | 0.7940517933336151 | 0.654854199500387 | 0.001 |
|
19 | 0.13720253109931946 | 0.2654192654192654 | 0.793669650812508 | 0.6522812524843972 | 0.001 |
|
20 | 0.13964051008224487 | 0.253984753984754 | 0.791502353390154 | 0.6515497507659908 | 0.001 |
|
21 | 0.13785456120967865 | 0.2577962577962578 | 0.7925025501530093 | 0.6542904488686327 | 0.001 |
|
22 | 0.13633865118026733 | 0.2661122661122661 | 0.7952276188864443 | 0.6524154901292529 | 0.001 |
|
23 | 0.13627886772155762 | 0.27096327096327094 | 0.7961679924728424 | 0.656651787807274 | 0.001 |
|
24 | 0.14012865722179413 | 0.2661122661122661 | 0.7871861324722778 | 0.6438900918479138 | 0.001 |
|
25 | 0.1359640210866928 | 0.27546777546777546 | 0.7960565795113589 | 0.6538094573584412 | 0.001 |
|
26 | 0.1370791494846344 | 0.2692307692307692 | 0.7942222975262623 | 0.6407905722004358 | 0.001 |
|
27 | 0.13669614493846893 | 0.2654192654192654 | 0.7902460077686664 | 0.6469565906332285 | 0.001 |
|
28 | 0.1371130496263504 | 0.26888426888426886 | 0.7912144926283021 | 0.642689033704319 | 0.001 |
|
29 | 0.13781629502773285 | 0.2692307692307692 | 0.7944120277694962 | 0.6484600603294314 | 0.001 |
|
30 | 0.13641151785850525 | 0.26507276507276506 | 0.7938241064573914 | 0.6472439075890195 | 0.001 |
|
31 | 0.13565559685230255 | 0.2747747747747748 | 0.7999161777032691 | 0.6533472550118105 | 0.001 |
|
32 | 0.137930765748024 | 0.2664587664587665 | 0.7928646379853095 | 0.662032330499469 | 0.001 |
|
33 | 0.13557712733745575 | 0.273042273042273 | 0.7989514185446704 | 0.6722007856831675 | 0.001 |
|
34 | 0.1347290426492691 | 0.273042273042273 | 0.7966670917825107 | 0.670590685863264 | 0.001 |
|
35 | 0.13544337451457977 | 0.2772002772002772 | 0.7946646145953571 | 0.6482708127714739 | 0.001 |
|
36 | 0.13763058185577393 | 0.25848925848925847 | 0.7927604900328681 | 0.6552995006011981 | 0.001 |
|
37 | 0.13456694781780243 | 0.2747747747747748 | 0.7992204380799051 | 0.6680976075122991 | 0.001 |
|
38 | 0.13784632086753845 | 0.27165627165627165 | 0.7889066758966815 | 0.6543467314054483 | 0.001 |
|
39 | 0.13671767711639404 | 0.2664587664587665 | 0.7965357098029371 | 0.6627442989440849 | 0.001 |
|
40 | 0.13730555772781372 | 0.27373527373527373 | 0.8004978220286246 | 0.670153584497431 | 0.001 |
|
41 | 0.13770104944705963 | 0.26576576576576577 | 0.7942296990711015 | 0.6610276871242879 | 0.001 |
|
42 | 0.13536451756954193 | 0.28101178101178104 | 0.8001525876319246 | 0.6705886094654014 | 0.001 |
|
43 | 0.13665379583835602 | 0.26507276507276506 | 0.8000498525196295 | 0.6619628883017729 | 0.001 |
|
44 | 0.12908011674880981 | 0.2869022869022869 | 0.808658516161447 | 0.6825865030851337 | 0.0001 |
|
45 | 0.12761357426643372 | 0.29972279972279975 | 0.811512367788968 | 0.6938587241702103 | 0.0001 |
|
46 | 0.12698666751384735 | 0.2959112959112959 | 0.8103163511624953 | 0.6856377454961721 | 0.0001 |
|
47 | 0.12690682709217072 | 0.2972972972972973 | 0.8124920976103174 | 0.6942647446672258 | 0.0001 |
|
48 | 0.12617328763008118 | 0.29799029799029797 | 0.8131711409395973 | 0.694151320978192 | 0.0001 |
|
49 | 0.1263018250465393 | 0.2966042966042966 | 0.8147346514047868 | 0.6956458198072734 | 0.0001 |
|
50 | 0.1258096992969513 | 0.2927927927927928 | 0.8153475224476222 | 0.7006577033751422 | 0.0001 |
|
51 | 0.12573884427547455 | 0.2972972972972973 | 0.8151571934207786 | 0.6994505755010588 | 0.0001 |
|
52 | 0.12501972913742065 | 0.2972972972972973 | 0.8134649455833967 | 0.6974514657531053 | 0.0001 |
|
53 | 0.12481856346130371 | 0.2948717948717949 | 0.8132960287301124 | 0.6962280886309719 | 0.0001 |
|
54 | 0.12473563104867935 | 0.30180180180180183 | 0.8143470573377115 | 0.6980743235485474 | 0.0001 |
|
55 | 0.12453257292509079 | 0.30076230076230076 | 0.8165587111775452 | 0.7020497284253308 | 0.0001 |
|
56 | 0.12440259009599686 | 0.3011088011088011 | 0.8185497191939213 | 0.7041152638460181 | 0.0001 |
|
57 | 0.12393573671579361 | 0.3004158004158004 | 0.8162207357859533 | 0.6984123654445143 | 0.0001 |
|
58 | 0.12355069816112518 | 0.30006930006930005 | 0.8171478565179352 | 0.7041206694443728 | 0.0001 |
|
59 | 0.1237163171172142 | 0.3049203049203049 | 0.8158932617269447 | 0.701908769020469 | 0.0001 |
|
60 | 0.12339853495359421 | 0.29902979902979904 | 0.8152564590468943 | 0.7008492179245241 | 0.0001 |
|
61 | 0.12294851988554001 | 0.3024948024948025 | 0.8188720173535793 | 0.7083200505706103 | 0.0001 |
|
62 | 0.12270853668451309 | 0.30284130284130284 | 0.8166017506386899 | 0.7054890147149661 | 0.0001 |
|
63 | 0.12301415950059891 | 0.3038808038808039 | 0.8176490288010717 | 0.7105833307429198 | 0.0001 |
|
64 | 0.12328237295150757 | 0.3049203049203049 | 0.8191759178412541 | 0.7085844813380441 | 0.0001 |
|
65 | 0.12309526652097702 | 0.3049203049203049 | 0.8187567612548888 | 0.7103887558295827 | 0.0001 |
|
66 | 0.12194398790597916 | 0.30284130284130284 | 0.8186407442947141 | 0.7061406642055487 | 0.0001 |
|
67 | 0.12292120605707169 | 0.3042273042273042 | 0.8196775527077305 | 0.7154558287425048 | 0.0001 |
|
68 | 0.12254418432712555 | 0.30803880803880807 | 0.8209686046990085 | 0.7153434473934246 | 0.0001 |
|
69 | 0.12215162813663483 | 0.3031878031878032 | 0.8195983668027664 | 0.7101570111652898 | 0.0001 |
|
70 | 0.12227334082126617 | 0.30838530838530837 | 0.8184682603033231 | 0.7109091736321397 | 0.0001 |
|
71 | 0.12237659096717834 | 0.3076923076923077 | 0.8170385739086251 | 0.7120407268503043 | 0.0001 |
|
72 | 0.1220996230840683 | 0.3063063063063063 | 0.8203632727878687 | 0.7203981522602361 | 0.0001 |
|
73 | 0.12169401347637177 | 0.3087318087318087 | 0.8198457369189076 | 0.7144193511981376 | 1e-05 |
|
74 | 0.12149834632873535 | 0.30665280665280664 | 0.8190452070406484 | 0.7124121424308173 | 1e-05 |
|
75 | 0.12120900303125381 | 0.30561330561330563 | 0.8208643316893754 | 0.7145366354361308 | 1e-05 |
|
76 | 0.1215985044836998 | 0.30803880803880807 | 0.8218541121766927 | 0.7191205487713891 | 1e-05 |
|
77 | 0.1214083805680275 | 0.31323631323631324 | 0.8236983547367989 | 0.7202749659896155 | 1e-05 |
|
78 | 0.12110316008329391 | 0.3097713097713098 | 0.8222591362126246 | 0.7168480610158249 | 1e-05 |
|
79 | 0.12149946391582489 | 0.30665280665280664 | 0.8202977563430488 | 0.7160500850094047 | 1e-05 |
|
80 | 0.121590256690979 | 0.30734580734580735 | 0.8219257062844905 | 0.7150848378423871 | 1e-05 |
|
81 | 0.12097962200641632 | 0.3115038115038115 | 0.8216162121591194 | 0.7187103786018064 | 1e-05 |
|
82 | 0.12082336097955704 | 0.30942480942480943 | 0.821175978238125 | 0.7156786549052798 | 1e-05 |
|
83 | 0.12147542089223862 | 0.30006930006930005 | 0.8180206046275968 | 0.7102312532643303 | 1e-05 |
|
84 | 0.12100570648908615 | 0.31185031185031187 | 0.8215978053038491 | 0.7195842513107142 | 1e-05 |
|
85 | 0.1208326444029808 | 0.31011781011781014 | 0.8233587533156498 | 0.7201395616901511 | 1e-05 |
|
86 | 0.1210438683629036 | 0.30942480942480943 | 0.8218151540383014 | 0.7215167678270465 | 1e-05 |
|
87 | 0.1212099939584732 | 0.3087318087318087 | 0.8207271207689094 | 0.7141558876633265 | 1e-05 |
|
88 | 0.12096676975488663 | 0.31011781011781014 | 0.8223957468017943 | 0.7124615854591595 | 1e-05 |
|
89 | 0.12144902348518372 | 0.3121968121968122 | 0.8240642149234173 | 0.7249978662662346 | 1.0000000000000002e-06 |
|
90 | 0.12115956842899323 | 0.31046431046431044 | 0.8233893154847453 | 0.7198781344667567 | 1.0000000000000002e-06 |
|
91 | 0.1208055168390274 | 0.3097713097713098 | 0.8212459126351974 | 0.7159843095789674 | 1.0000000000000002e-06 |
|
92 | 0.12069901078939438 | 0.30734580734580735 | 0.8223893065998329 | 0.7144036362020703 | 1.0000000000000002e-06 |
|
93 | 0.12093978375196457 | 0.30803880803880807 | 0.8226574468966088 | 0.7189178649032102 | 1.0000000000000002e-06 |
|
94 | 0.12092197686433792 | 0.3097713097713098 | 0.8223438666334908 | 0.7187657914933285 | 1.0000000000000002e-06 |
|
95 | 0.1206900030374527 | 0.30942480942480943 | 0.8221934621968021 | 0.7127077698746517 | 1.0000000000000002e-06 |
|
96 | 0.12142115086317062 | 0.30665280665280664 | 0.8218438538205979 | 0.7160309422692305 | 1.0000000000000002e-06 |
|
97 | 0.12264719605445862 | 0.30942480942480943 | 0.8208711661575798 | 0.71586766610014 | 1.0000000000000002e-06 |
|
98 | 0.12095578759908676 | 0.31185031185031187 | 0.8224561403508771 | 0.7190138873820752 | 1.0000000000000002e-06 |
|
99 | 0.12075632065534592 | 0.3097713097713098 | 0.821403230518803 | 0.7177436878101541 | 1.0000000000000002e-07 |
|
100 | 0.12078335881233215 | 0.3108108108108108 | 0.8218776194467728 | 0.7191112023643382 | 1.0000000000000002e-07 |
|
101 | 0.12071150541305542 | 0.3097713097713098 | 0.8230599775551769 | 0.7199208624613478 | 1.0000000000000002e-07 |
|
102 | 0.12102664262056351 | 0.31011781011781014 | 0.821560093739538 | 0.7181176324357539 | 1.0000000000000002e-07 |
|
103 | 0.12072332948446274 | 0.31115731115731116 | 0.8218559116391932 | 0.7156251632807489 | 1.0000000000000002e-07 |
|
104 | 0.12122868001461029 | 0.3090783090783091 | 0.8214226220223222 | 0.7151217785983346 | 1.0000000000000002e-07 |
|
105 | 0.12081456929445267 | 0.30838530838530837 | 0.8216449497883642 | 0.7175066761763569 | 1.0000000000000004e-08 |
|
|
|
|
|
--- |
|
|
|
# CO2 Emissions |
|
|
|
The estimated CO2 emissions for training this model are documented below: |
|
|
|
- **Emissions**: 0.5035923822963007 grams of CO2 |
|
- **Source**: Code Carbon |
|
- **Training Type**: fine-tuning |
|
- **Geographical Location**: Brest, France |
|
- **Hardware Used**: NVIDIA Tesla V100 PCIe 32 Go |
|
|
|
|
|
--- |
|
|
|
# Framework Versions |
|
|
|
- **Transformers**: 4.41.1 |
|
- **Pytorch**: 2.3.0+cu121 |
|
- **Datasets**: 2.19.1 |
|
- **Tokenizers**: 0.19.1 |
|
|
|
|
